114 CARNEGIE INSTITUTION OF WASHINGTON. 



the preliminary trials) constant through the training or is it modified by the 

 training? When the controls were grouped according to their preliminary 

 performance, the averages for the successive sets of 20 trials (2 days) showed 

 very slight relationship to the preliminary records. When the tests are simi- 

 larly grouped, their preliminary performance is so closely related to the sub- 

 sequent trials that a good prediction of the later behavior can be made from 

 the preliminary performance. The statistical study of this point shows 

 that the preliminary performance of the controls is very slightly correlated 

 with the average of each successive set of 20 trials, whereas comparable 

 correlation coefficients for the tests are high. This study has revealed the 

 most striking difference between the tests and controls that has been brought 

 out by the multiple-choice apparatus. The reactions of the controls are 

 ahnost entirely changed by the introduction of the training, while the reac- 

 tions of the tests remain about the same as in the preliminary trials. This 

 may be a difference in responsiveness to changes in the situation, the tests 

 being less responsive to the changed circumstances than the controls. In 

 view of the above, the least objectionable way of making the final compari- 

 sons of averages of tests and controls appears to be to eliminate all rats with 

 preliminary records beyond certain limits (for instance, with more than 12 

 or less than 3 correct trials in the preliminary trials). This removes the 

 difference in the averages for the preliminary trials. When this is done the 

 graphs of the averages show the tests making fewer correct choices in each 

 set of 20 trials after the first one. As the differences increase they become 

 statistically significant. 



"Behavior in maze: In the Year Book for 1920 the maze-behavior of the 

 children and grandchildren of alcoholized rats was reported. The children, 

 at that time, subject to correction, have been rechecked through the co- 

 operation of Dr. Harris. The great mass of the statistical work herein 

 reported, however, has been carried out in this department with the assist- 

 ance of Mr. L. H. Snyder and Miss Vicari. 



"The behavior data for the original treated rats and treated rats from 

 treated parents have been tabulated and the summaries are ready for pub- 

 lication. 71 original treated rats (with 67 controls) and 8 treated rats from 

 treated parents (with 20 controls) were given their trials in the maze each 

 day before the alcohol treatment. 



"In general, the tests took more time than the controls; the offspring of 

 treated rats show less difference than any other generation. The other 

 three generations (1, treated; 2, treated from treated parents; and 3, un- 

 treated from untreated parents and treated grandparents) show more or less 

 the same general differences between the tests and controls. In other words, 

 this difference is neither dependent upon the presence of alcohol itself, nor 

 is it intensified by the treatment of two successive generations. This result 

 is similar to that given by the size of the litters produced in the different 

 generations." 



THE SIGNIFICANCE AND CONTROL OF SEX. 

 The Fundamental Nature of Sex. 



The phenomenon of two sexes runs through the world of organisms 

 with a constancy that is hardly found in any other phenomenon, 

 yet with sporadic apparent exceptions which have attracted the atten- 

 tion of biologists from an early time. A phenomenon of such univer- 

 sality among organisms, and yet not recognized outside of organisms, 

 must have some deep biological significance. Though in the past 



